Surya Text-Line Detection — MLX

MLX port of surya's EfficientViTForSemanticSegmentation text-line detector (EfficientViT-L backbone + SegFormer head). Box-F1 vs the original ~0.99-1.00 on sample documents (see bench/RESULTS.md). Loads torch-free via this project's surya2_mlx.detection.

Changes from the original

Weights converted to MLX (NHWC) format. No fine-tuning. Architecture and source: https://github.com/datalab-to/surya (Apache-2.0 code; weights under datalab terms).

Downloads last month
19
Safetensors
Model size
38.4M params
Tensor type
F32
·
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support